Statistics for Meta-learning adaptive intrinsic reward weighting for curiosity-driven reinforcement learning